You are on page 1of 4

A Verilog-A model for the design of devices for

fluorescence life-time measurement with CMOS


SPADs

A. Dieguez, o. Alonso, E. Vilella, A. Vila


Department of Electronics
University of Barcelona
Carrer Marti Franques, 1. 08028-Barcelona, Spain
adieguez@el.ub.edu

Abstract- Point of care devices for the early diagnosis of that affect fluorescence intensity.
current relavant diseases of our society require novel Nowadays, fluorescence measurements are still performed
miniature and very effective devices. This contribution by fluorescence microscopes, which are bulky and expensive.
focuses on the description of behavioral models aimed to Future personalized medicine needs small and effective
develop custom readout Application-Specific Integrated diagnosis tools, including diagnosis by fluorescence. Quite
Circuits (ASICs) for the measurement of Fluorescence recently, the use of on-chip fluorescence and life-time
and Fluorescence life-time of target substances. The well detection with CMOS SPADs has been addressed by several
suited for the ASIC development Verilog-A behavioral research groups. In order to provide low cost, fully integrated
description language is used to describe the sensor, an solutions, fulfilling the necessary constraints in terms of
ultra sensitive Single Photon Avalanche Diode (SPAD) important parameters like power, significant progress is still
fabricated in a conventional CMOS process, and the needed. Having useful models for the SPAD sensor and the
fluorescence signal. fluorescence signal is of relevance for the development of the
new instrumentation and is also the aim of this contribution.
Index Terms-Verilog-A, behavioral, SPAD, Avalanche
Behavioral models allow the optimization of the design
photodiode, CMOS, fluorescence, molecular, diagnosis, life-time, choice at the simulation level, thus saving costs and time.
FLlM. Moreover, they allow to explore different architectures during
the development phase, which is of paramount importance to
I. INTRODUCTION
succeed in integration.
This contribution describes the key simulation models used,
Single-photon Avalanche Diodes (SPADs) are optical sensors which are based on the high level mixed-signal modeling
that can detect single photons. Due to the weakness of the language Verilog-A. Verilog-A allows the description of the
signal in biological experiments, SPADs offer great potential electrical and non-electrical parts of the system, allowing a
as highly sensitive biosensors. Moreover, SPAD devices can whole integration in a standard simulation environment
be fabricated in conventional CMOS processes, thus commonly used to develop integrated circuits. The description
exploiting the advantages of microelectronic technologies, of the electronics can be done at the maximum level (post­
i.e., low cost, low power and high density of integration. layout) and the sensor and the incoming signal in Verilog-A.
Currently, CMOS SPADs are gaining interest in many single­ The sensor model can emulate the static and dynamic
photon detection applications. behaviors, and the main statistical effects of an SPAD. The
In this paper, we focus on the application of SPADs to the incoming signal is described as a set of photons reaching the
measurement of the Fluorescence life-time of excited sensor randomly in time, following an exponential decay. The
molecules. Fluorescence methods applied to medical sciences simulations reveal the characteristics required to track the
exploit the phenomenon of fluorescence to analyze fluorescence decay of different substances.
biomolecular details of living tissue. In recent years, they Section II overviews a general set-up for fluorescence
have become the key tool of many optical bioassays for the measurements and the SPAD detector. The Verilog-A models
instantaneous diagnosis and monitoring of a large number of for the SPAD and the fluorescence signal are presented in
diseases, including cancer, cardiorespiratory, Section III. In Section IV we describe an example of
neurodegenerative and immunological diseases. In particular, integration. Finally, the conclusions are given.
the measurement of the life-time has the advantage of being
insensitive to the fluorophore concentration and other factors
978-1-4673-9184-9/15/$31.00 ©lOIS IEEE
II. FLUORESCENCE MEASUREMENTS WITH SPADs v••

VB. + Vov

The principle of operation of SPADs is based on the K


"
'"

avalanche multiplication phenomenon. Although the first -


" -

.!l-"
studies on this phenomenon in p-n junctions started in the A " "
0 E

1960s at the Shockley laboratory [I, 2], it was not until the $! 0
"

1990s that solid-state avalanche detectors became available in Vs

a CMOS compatible process [3]. However, only the


monolithical integration of an SPAD sensor and the front-end
electronics on a single CMOS die in 2003 [4] opened the way
to commercial applications. Currently, one of the applications
with the highest potential is Fluorescence Lifetime Imaging
Microscopy (FUM) measurements [5-7].
A typical fluorescence decay measurement setup consists of
Figure 2: Typical SPAD pixel. The diode is biased VOY (overvoltage)
a pulsed light source, a sensitive detector (the SPAD) and an above the breakdown voltage. Vbias controls the SPAD recharge process after
optical system. For time-resolved fluorescence analysis, a sub­ the avalanche. The sensor is read by a CMOS inverter formed by MNJ and
nanosecond pulsed light source and usually Time Correlated MP4- The signals RS T and INH are used to control the time that the sensor is
active (gating of the sensor). The pixel is reset by RST and disabled with INH.
Single Photon Counting (TCSPC) are used. In TCSPC, the
It is only active after the RST iflNH is high.
time of arrival of the photons after an excitation pulse is
measured and classified to reconstruct the exponential decay
of the fluorescence. Figure I illustrates the TCSPC process.

� Time /

time 0

time 1 1\ Figure 3. Cross section of one SPAD fabricated with the standard HV-AMS
0.35 J.lm CMOS technology. The deep p-tub in the edge of the p+/deep ntub
junction avoid premature breakdown.
1\
1\
III. BEHAVIORAL DESCRIPTIONS

1\
The behavior of an SPAD sensor is not described by
technology models. It is a p-n junction biased some Vav
above the breakdown voltage, thus working in the Geiger­
time n
1\ mode. Several works have presented electrical models of
reconstructed histogram SPADs, including behavioral descriptions [9] representing
-
even the statistics of noise in these devices. For the purpose of
-

simulating the fluorescence detection a simple development is


�-
-
necessary. The model used for the SPAD is presented in

-
figure 4, after [ 10], for a SPAD fabricated in a 180 nm
process. It is a piecewise linear model describing the static
Figure I: TCSPC process. Photons are detected at different excitation behavior of the SPAD, that avoids convergence problems by
times. The time of arrival is used to reconstruct the decay of the fluorescence. using the defined functions f 1- f4.
The signal, as illustrated in figure 1 for a TCSPC
As commented, an SPAD detector can be fabricated in a measurement, is generated with the code shown, in portions,
conventional CMOS process. A typical pixel being part of an in figure 5. The model considers a fixed number of events
array is as presented in figure 2 [8]. The SPAD device is which correspond to the arrival of photons to the sensor. This
constructed in such a way that the electric field at the edges of reduces substantially the calculations in a simulation. The time
the p-n junction is lower to avoid premature edge breakdown. for each event is calculated in basis of an exponential or
The sensitive part is at the bottom of the structure (figure 3).
module APD_model_p_plus_in_nwell(cathode,anode,photon); module rand_event_stream (reset, vout);

inout cathode,anode,photon; output vout;


electrical cathode,anode,photon; input reset;
electrical vout, reset;
II Parameter definition
II Parameter definition
parameter real Vbd = 22.3; II depends on process and structure
parameter real R_d = 100.0; II SPAD internal R II main code
parameter real R_turnoff = 250.0; II Turn-off resistance
parameter real Uatchup = 20.0E-6; II Latchup current analog begin
parameter real Cs = 1O.0E-12; II Reverse saturation current @ (initial_step) begin
parameter real n = IE-5; II Scale factor to avoid overflow Illnitial definitions
parameter real C_d = 134.73e-15; II SPAD capacitance
II calculate exponential decay: event time is exponentially
real V_d,Yov; decreasing, P= exp(-i*ln/tdecay)
real C_p_d;
real Q_d; for ( i=l; i <= density; i=i+l ) begin
real avalanche; probability = 1.0*i/density;
real f1,f2,f3,f4,INL; timeofevent[i-I] = (-1.0) * tdecay * In (probability);
end
analog begin end

V_d = V(cathode) - V(anode); II


Vov = V_d - Vbd;
II After a reset signal the event is get from the table
@(initial_step)
begin @ (cross(V(reset)-(vlogic_high-vlogic_low)*0.5, I)) begin
avalanche = 0.0; II Initital conditions (quiescence)
end II select the index for the time in the list
eventselect = $dist_uniform(seed, 0, queuesize-l);
if (V(photon) > 0.99) event = $abstime + timeofevent[eventselect] ;
begin bit = 0;
avalanche = 1.0; II To mimic avalanche triggering vout_val = vlogic_low;
end

fl = 1_s; II the event is eliminated from the list


f2=(VovlR_turnofl)-((R_d - R_turnofl)*Uatchup)lR_turnoff;
f3 = abs(Vov/R_d) + U;
f4 = n*ln(exp(fl/n) + exp(J2/n)); II the signal to the event is generated and feed to the environment
INL=avalanche*(-n*In(exp(-f3/n)+exp(-f4/n)))+! (avalanche)*Cs;
@ ( timer( event )) bit = I;
if (INL < Uatchup) @ ( timer( event + tperiod)) bit = 0;
begin vout_val = (vlogic_high - vlogic_low) * bit + vlogic_low;
avalanche = 0.0; V(vout) <+ transition(vout_val,tdel,trise,tfall);
end
end
Q_d = C_d*(V(cathode,anode));
endmodule
1(cathode,anode) <+ ddt(Q_d);
I(cathode,anode) <+ INL;

end
endmodule

Figure 4: Verilog-A code of an avalanche photodiode, considering a 180 Figure 5: Verilog-A code for the photon generation in a FUM TCSPC
nm process. experiment.

multi-exponential probability, depending on the target Figure 6 shows the result of the application of these models
substance to simulate. A table stores the time of arrival of the to the circuit presented in figure 2. An event is generated some
different events (photons) after firing a start signal (pulse of time after the reset signal, as expected. The anode of the
light). A section of the code randomly takes one of the SPAD goes from VSS after the reset, to the overvoltage at the
generated events and reduces the table size, eliminating such event time. The readout, consisting basically in a CMOS
event from the table. After the complete simulation the inverter, is going from VSS to VDD ( 1.8 V).
histogram must be reconstructed.
REFERENCES

[ 1] R.H. Haitz, Model for the electrical behavior of

--------'-
microplasma, J. App!. Phys., vo!' 35, pp. 1370- 1376,
May 1964.
[2] R.H. Haitz, Mechanisms contributing to the noise pulse
rate of avalanche diodes, J. App!. Phys., vo!' 36, pp.
3 123-3 13 1, Oct. 1965.
[3] A.C. Giudice et aI., A CMOS compatible single-photon
avalanche diode, in Proc. 32nd European Solid-State
Device Research Conference, Firenze, 2002.
[4] A. Rochas, M. Gani, B. Furrer, P.A. Besse, R.S. Popovic,
and G. Ribordy, Single photon detector fabricated in a
complementary metal-oxide-semiconductor high­
voltage-technology, Rev. Sci. Instrum., vol. 74, pp.
3263-3270, Apr. 2003.
[5] D. Stoppa, D. Mosconi, L. Pancheri, and L. Gonzo,
Single-photon avalanche diode CMOS sensor for time­
resolved fluorescence measurements, IEEE Sens. J., vol.
9, pp. 1084-1090, Sep. 2009.
Figure 6. Simulation in the Virtuoso Cadence design environment of the event
generation and the SPAD output, as inputs to the readout presented in Figure [6] B. Rae, J. Yang, J. McKendry, Z. Gong, D.R. Renshaw,
2. The 'event' signal induces an SPAD firing (see the anode going to J.M. Girkin, E. Gu, M.D. Dawson, R.K. Henderson,
Overvoltage after the initial RST). The PIX signal is the output of the readout. A Vertically Integrated CMOS Microsystem for Time­
IV. CONCLUSION
Resolved Fluorescence Analysis. IEEE Trans. Biomed.
Circuits and Systems 4(6), pp. 437-444 (20 10).
[7] D. Tyndall , B. Rae , D. Li , J. Richardson , J. Arlt and R.
The behavioral models of SPAD detectors and the signal in a
Henderson, A 100 Mphoton/s time-resolved mini-silicon
fluorescence-decay measurement have been presented. These
photomultiplier with on-chip fluorescence lifetime
are used in the ASIC development environment. The
estimation in 0. 13um CMOS imaging technology, Proc.
application of these models in the Cadence environment has
Int. Solid State Circuits Conf. Dig. Tech. Papers, pp. 122
been shown. This is a step towards the development of new
- 124, 20 12.
diagnosis tools in the field of personalized medicine.
[8] E. Vilella, O. Alonso, A. Montiel, A. Vila, and A.
Dieguez, A low-noise time-gated single-photon detector
ACKNOWLEDGMENT in a HV-CMOS technology for triggered imaging, Sens.
Actuators A Phys. 20 1, 342-35 1 (20 13).
This work has been partially supported by ACC 16 in
Catalonia through the Project 'Portable device for molecular [9] G. Giustolisi, R. Mita and G. Palumbo, Behavioral
diagnosis' coded VALTEC13- 1-0020-00, and by the Spanish modeling of statistical phenomena of single-photon
Program for Particle Physics through the project 'Development avalanche diodes. Int. J. Circ. Theor. and App!. 40, pp.
of new detectors and physics studies for future linear colliders' , 66 1-679, 20 12.
coded FPA20 13-48387.4. [ 10] Mita R, Palumbo G, Fallica PG. Accurate model for
single-photon avalanche diodes. lET Circuits, Devices
and Systems 2(2), pp. 207-2 12, 2008.

You might also like